(Remote)

Talk Title: Computer Vision and the Metaverse

Abstract: Computer vision technology has advanced tremendously in recent years, but it is of limited use by itself. It’s full value is only realized when it becomes well integrated into larger user solutions in the real or virtual world, involving training, inference, performance evaluation or other aspects. Custom integrations are always possible, but if we can integrate computer vision into standardized infrastructure, it will become vastly simpler to deploy.

Our best example of a universal platform for integrating varied information technologies together is the world-wide web. Unfortunately the current web has its feet solidly planted in the 2D world. A variety of efforts have been made to add 3D capabilities to the web, but these have failed to address the central issue. 3D is fundamentally harder than 2D. So in order to create the solution we really want, we should start with the needs and requirements of 3D, and then integrate 2D into a proper 3D web instead of the other way around. Here, we explore what that proper 3D web should look like, and how it can act as the foundation of the metaverse. The full vision remains to be realized, but we will show some of the concrete steps that NVIDIA has already taken in this direction with the Omniverse platform.

Bio: Michael Kass is a senior distinguished engineer at NVIDIA and the architect of NVIDIA Omniverse, NVIDIA's platform for collaborative 3D content creation and digital twins. In 2005, Kass received a Scientific and Technical Academy Award for “pioneering work in physically-based computer-generated techniques used to simulate realistic cloth in motion pictures.” In 2009, he received the ACM Computer Graphics Achievement Award for "his extensive and significant contributions to computer graphics, ranging from image processing to animation to modeling and in particular for his introduction of optimization techniques as a fundamental tool in graphics." And in 2017, the ACM honored him as an ACM Fellow “for contributions to computer vision and computer graphics, particularly optimization and simulation.” Kass has been granted over 30 U.S. patents, and was honored in 2018 as Inventor of the Year by the NY Intellectual Property Law Association. Before switching to computer graphics, he had an extensive career in computer vision. His Helmholtz-award winning computer vision paper “Snakes: Active contour models” is one of the most cited papers in computer science with over 25k citations. Kass holds a B.A. from Princeton, an M.S. from M.I.T. and a Ph.D. from Stanford.